Interview Build Planner — 16 Week
16-Week Interview Build Plan · ~2 hrs/workday, 3–4 hrs/weekend · deep learning + NeetCode 250 + 7 projects · ready to apply after Week 16
How to use
Priority: DSA → System Design (+CS) → ML/DL/LLM/VLM → MLOps. 16 weeks is the learn-properly pace, not a crash. Each week is a page below; tick items as you go (checkboxes reset on reload — track real progress in LeetCode + your tracker).
Daily (~2 hrs): 75 min DSA (timed, no AI autocomplete) + 10 min syntax typing + 35 min second block (alternate System Design / ML). Weekends: long DSA set + project + 1 system-design practice out loud.
Golden rules: talk out loud · after a working solution spend 60s on complexity · over the time cap → redo from scratch 48h later · describe the mechanic if a pattern name escapes you.
Weekly rhythm (assumption: Day 1 = Monday)
Each week assumes a Monday start, five workdays and a weekend:
- Mon–Fri (workdays, ~2 hrs): 75 min DSA (timed) + 10 min syntax typing + 35 min second block. Second block alternates Mon/Wed = System Design + CS, Tue/Fri = ML/DL/LLM/VLM, Thu = MLOps / catch-up.
- Sat–Sun (~3–4 hrs): one long DSA set (the week's harder problems + any leftovers), project work, and one system-design practice out loud.
- Pacing: ~19–20 problems/week ≈ 4 per workday, with the hardest saved for the weekend when you have more time.
Shift the start day if your week begins differently — the structure (5 learning days + a heavier weekend) is what matters, not the calendar.
Shape of the 16 weeks
Weeks 1–13 — build the foundation. All 250 NeetCode problems by pattern, plus a deep second-block curriculum that progresses each week:
- System Design + CS: OS → networking → DBMS → system-design core → distributed systems → ML system design → inference optimization → MLOps.
- ML / DL / LLM / VLM: foundations → regression → ensembles → neural nets & backprop → CNNs → sequence models → Transformers → LLM fundamentals → fine-tuning → VLMs → RAG → agents & evaluation.
Weeks 14–15 — convert to performance. Mock loops (coding + system design + behavioral), weak-area redo, resume + portfolio, and begin applying.
Week 16 — consolidate & apply broadly into the next hiring wave.
Timing tightens over the plan: Week 1 Medium ≤ 35m → Week 12+ Medium ≤ 22m, Hard ≤ 40m, talking throughout.
Projects (6 core + 1 optional capstone, built mid-plan)
- P1 Recommender System (Wk 3–4) — Recall@K/NDCG, FastAPI, Docker. Fills recsys gap.
- P2 Multi-Model Inference Server + Benchmark (Wk 5–6) — latency/throughput, monitoring, CI. Your Triton/GPU-scheduling story, public.
- P3 RAG App + LLM Eval Pipeline (Wk 7–8) — FAISS retrieval, LLM-as-judge eval. Applied-Scientist / Turing fit.
- P4 VLM Fine-Tuning (Wk 9–10) — PaliGemma/BLIP/LLaVA. Your CV + VLM strength.
- P5 ML System Implementation (Wk 11–12) — feature store / embedding-serving with monitoring.
- P6 End-to-End MLOps Pipeline (Wk 13–14) — training → registry → CI/CD → serving → monitoring.
- P7 Capstone / Portfolio polish (Wk 15, optional) — writeup/blog + portfolio.
Only list what is actually on your public GitHub and defensible end to end.
Free resources
- DSA: NeetCode (free roadmap + YouTube), Striver / takeuforward.org, LeetCode
- Syntax: SyntaxCache, SpeedCoder, Anki
- System design: System Design Primer (GitHub), ByteByteGo, Gaurav Sen, Hello Interview, Exponent, systemdesign.io, Google SRE book
- CS fundamentals: OSTEP (OS), Neso Academy (OS/DBMS/Networks), use-the-index-luke (DB indexing)
- ML/DL: StatQuest, 3Blue1Brown, Illustrated Transformer, Chip Huyen "ML Interviews"
- LLM/VLM: HuggingFace NLP course, PEFT docs, CLIP/Transformer papers, HF blog
- MLOps: Made With ML, Google MLOps guides, NVIDIA Triton/TensorRT docs